"Empowering Data-driven AI by Argumentation and Persuasion" chair
In 2022, we investigated theoretical foundations of explainability, a hot topic in AI. In , we laid the foundations of explanation models by proposing key axioms, i.e., desirable properties they would satisfy, and characterized various families of models satisfying subsets of axioms. In , we analysed existing models against the axioms, highlighted their properties links and shortcomings. In , we proposed novel explanation models that satisfy the axioms while overcoming the shortcomings from the literature. Finally, we proposed in  a hybrid approach for ensemble classifiers. It combines machine learning models and nonmonotonic logics, namely argumentation-based logics. We show that the novel ensemble classifiers guarantee desirable properties including explainability, compliance to knowledge, and a global compatibility of the rules they use for making predictions. An experimental study conducted in the healthcare domain shows that the hybrid approach competes with existing ensemble methods.
 L. Amgoud, J. Ben-Naim. Axiomatic Foundations of Explainability. In Proceedings of the International Joint Conference on Artificial Intelligence, IJCAI-22.
 L. Amgoud. Explaining Black-box Classifiers: Properties and Functions. International Journal of Approximate Reasoning, 2023.
 Leila Amgoud, Henri Trenquier, Philippe Muller. Argument-based Explanation Functions. In Proceedings of the International Conference on Autonomous Agents and Multi-Agent Systems, AAMAS-23.
 L. Amgoud, D. Doder, S. Versic. Evaluation of Argument Strength in Attack Graphs: Foundations and Semantics. Artificial Intelligence Journal, 2022.
 Leila Amgoud, Nadia Abchiche, Farida Zehraoui. Explainable Ensemble Classification Models Based on Argumentation. In Proceedings of the International Conference on Autonomous Agents and Multi-Agent Systems, AAMAS-23.